Task Ontology-Based Framework for Modeling Users' Activities for ...

Task Ontology-Based Framework for Modeling Users’ Activities for Mobile Service Navigation Munehiko Sasajima1, Yoshinobu Kitamura1, Takefumi Naganuma2, Shoji Kurakake2 and Riichiro Mizoguchi1 1

I.S.I.R., Osaka University 8-1 Mihogaoka, Ibaraki,

Osaka, 567-0047 Japan {msasa, kita, miz}@ei.sanken.osaka-u.ac.jp

Keywords Task ontology, mobile service, model of users' activity

1. INTRODUCTION We can get many kinds of mobile services via mobile handsets in Japan. According to an annual report by NTT DoCoMo [1], Japan's premier mobile communications company which manages mobile internet services and occupies about 58% of the market, we have more than 89,000 service sites today. On the other hand, a large number of services cause difficulties in searching, finding and selecting suitable services for consumer's needs. One of the reasons for the difficulties is that menus of the current mobile services are organized from the viewpoint of the domain. Users have to learn the menu system to access the services; hierarchical structure of the menu, relation between name of the category and services in the category. If a user wants to catch the last train, for example, such a domain-oriented menu will guide him/her as follows: “menu”, “latest information”, “traffic”, “train information”, “timetable” and “input start station name”. The user follows the menu and finally reaches the service that provides information about timetable. As this example shows, users have to translate “what they want to do” to “name of the menu” before getting mobile services they want. On the other hand, there is another type of menu which we call “task-oriented menu” [2]. The goal of this research is to realize a task-oriented menu system which enables more efficient mobile

Figure1. Framework for designing task oriented mobile service Demos and Posters of the 3rd European Semantic Web Conference (ESWC2006), Budva, Montenegro, 11th - 14th June, 2006

2

Network Laboratories, NTT DoCoMo, Inc. 3-5, Hikari-no-oka, Yokosuka City, Kanagawa, 239-8536 {naganuma, kurakake}@netlab.nttdocomo.co.jp service navigation. Result of the experiment shows that taskoriented menu is more efficient for retrieving information [2]. By task, we mean users’ problem solving activity in the real world. In the task oriented menu, the users seek for services by the name of the directory which represents a task they are involved in rather than the name of category which might be unfamiliar to them. Users select a menu that is most resemble to what they want to do; “get on the train”, “draw cash”, for example. It has potential of providing useful information for mobile service users quicker than that of a domain-oriented menu. Value of information depends on the quality of contextual information that contains. By quality, we mean whether the information corresponds to the needs of the users or not. Necessity of information lies in a task, not in a domain. You seek for information when you face a trouble, which is difficult to get over with knowledge at hand, on your way of achieving a task. Such a situation is the context and origin of the necessity for the information. With backgrounds discussed above, this article proposes a task ontology-based modeling framework for mobile service navigation. Fig.1 depicts the framework of our system where rectangles represent knowledge, rectangles with round corners represent modules and circles represent people. “Service providers” in Fig.1 design users’ activity models and mobile internet services through the interface module. Its output is the menu of the mobile internet services that is used by the “User of the mobile services”. Although the service providers usually have implicit business models about their own mobile services, they do not have generic task models for representing users’ activities. Generic models and task/domain ontologies which are designed by “Designer of Ontology” are referred to by the service providers to obtain concrete models by instantiating the generic models. “Designer of Ontology”, an important role of the authors, designs and maintains ontologies. The authors are specialists for building task ontologies [3], and have experiences of its application to the real world problem solving [5]. Although there are huge numbers of “tasks” in the real world, those have be solved by mobile handset users are small, since they are limited to daily-life tasks done out side home. Furthermore, to organize task concepts is easier than that of domain concepts, because it is independent of domain, is able to be decomposed into subtasks and has a generality in the abstract space. For example, a task concept “buy a ticket for a movie” consists of two task concepts, “buy something” and “receive service (Including model of queuing)”. Both concepts can be applied to modeling similar tasks in various domains. Task concept thus has a generality in its nature and hence we can organize its structure at a high level of abstraction.

(4)Describe ways to ralize prevention/solution task

transfer the method to others. Models of how to prevent/solve Part-of problems are described in three Move to the theme steps. Firstly, the designer Move by driving park Check my own car Have a car serviced Way describes possible obstacles for beforehand Bad weather Part-of ion on each partial task. For example, e nt i Check by self Way v t e Enter the theme c Pr Rapair tru Trouble of the Part-of park Solution bs the task “Drive from one’s Call repair service Way car Go to the parking O Part-of Repair by self Way home” has four possible Prepare a map Part-of Enjoy attractions n entio Buy map Way obstacles: “Trouble of the car”, Prepare a navigation Prev Part-of Prevention Missing route system Prepare portable Way tion Part-of “Incomplete route”, “Traffic Have meals Solu Way truc s Drive from b t navigation system io O n Way Part-of Seek for new one's home Obs jam” and “Fatigue of driving”. tru Prepare car navigation Way c ti routes Way Do Shopping on system Building models of possible Traffic jam Preventio Prepare a navigation n So Part-of lut Use navigation system system troubles is a unique feature ion Way Back to home Park at a Use a map Tired of Seek for new routes compared to previous researches parking lot P r driving even Planningmethod including a Way Sotilon route and a break point utio Make plan with rest like [2][6]. Since the most n Move by public Method to use transportation valuable mobile service is to Take a rest Way Way information site No vacant Move with a rent-aPrevention solve such problems that space P Reserve a parking r ev e car Pre-reservation method Way So ntion Prepare a method for lut occurred, our modeling method ion searching a parking Refer to the parking copes with obstacles on the spot. Way Way Search for another information site parking Next, the designer describes prevention/solution tasks for Figure2. An example of modeling process each of the obstacles (Fig.2(3)). Lastly, the ways to achieve the The authors are investigating reorganization of the mobile services tasks are modeled (Fig.2(4)). from the viewpoint of task. The approach based on task ontology Since models of obstacles are described for the task with fine enables service providers to describe users’ activity models in granularity, we can imagine more tasks for their prevention/ terms of generic task vocabulary which are detached from the solution than conventional modeling methods. In Fig.2, for domain model. Furthermore, specification of the modeling process example, to generate preventive ideas for obstacles about the task based on categorization of users’ activity provides them with “move to theme park” is more difficult than that for the “Park at guidelines. Based on the task ontology, our method contributes to the parking lot”, because the former task is abstract and contains building homogeneous and generic models． many obstacles according to its interpretation. Based on the decomposition of the task models, our method helps generation of 2. Framework new prevention/solution ideas. With a similar reason, generating Fig.2 represents a process of building a task-way representation ideas about the ways to achieve the prevention/solution tasks is model of users’ activity. A dotted rectangle with number (1) supported by our modeling method. corresponds to the basic model of users' activity. It is described by instantiating generic models and/or ontology. Description starts 3. Research Status from the task at the level of coarse granularity. Next, ways to The authors are currently building task ontologies and models on achieve the task are linked, and each of the ways is decomposed the proposed framework. Task model of the activities related to into a sequence of sub-tasks. Our “way” is similar to “method” of the theme park has been building at the high level of abstraction, CommonKADS [6] or “how to bundle” of the Business Process and we plan to apply the ontology to other domains (cf. Chap1). Handbook [7] to some extent. Following this process, task of the At the same time, we are conducting several experiments to coarse granularity is decomposed into sub-tasks via a few ways. evaluate the proposed framework in terms of (1) supports for The area with number (1) represents that a task “Move to a theme generation of ideas about mobile services (2) relationship between park” is achieved by three ways. One of the ways “Move by given ontological information and the quality of the user models driving one’s own car” is decomposed to three sub-tasks such as output (3) efficiency of the modeling with our guideline, and so on. “Go to the parking lot”, “Drive from one’s home” and “Park the car at the parking lot”. 4. REFERENCES An important guideline in this framework is that the model of [1] http://www.nttdocomo.co.jp/english/corporate/investor_relations/refe renc/annual_e.html daily activity is described based on the observation of physical [2] T.Naganuma and S. Kurakake., Task Knowledge Based Retrieval for activity on the spot. Cognitive activities such as “plan to move (2)Describe possible obstacles for each partial task.

(1)Describe user activity that are observed

(3)Describe tasks that prevent/solve each obsacle.

Ob str u

cti o

n

Way of Playing:: Play at the Theme-park

Ob

str

n tio uc

n tio uc str Ob

more efficiently” or “learn traffic information beforehand” are not described in the model. The guideline and modeling process based on decomposition of the task contribute to making modeling process easier and output models more objective. To realize a coherent task-oriented menu structure with a large scale problem, transfer of the modeling technology is important. If we allow modeling non-observable activities, quality of output models varies according to the skill of each model builder. The less knowledge about the task and/or domain he/she has, the worse output models will become. In such a case, process of modeling becomes implicit and we cannot

[3] [4] [5] [6] [7]

Services Relevant to Mobile User’s Activity, In Proceedings of the ISWC2005, 959--973, 2005. Mizoguchi, R., Vanwelkenhuysen, J., and Ikeda, M., Task Ontology for Reuse of Problem Solving Knowledge, In Proc. of KB\&KS '95, The Netherlands, 46--59. 1995. http://www.hozo.jp Kitamura, Y., Kashiwase, M. Fuse, M., Mizoguchi, R., Deployment of an Ontological Framework of Functional Design Knowledge, Advanced Engineering Informatics, 2004:18(2):115-127 . 2004. Schreiber, G., Akkermans, H., Anjewierden, A., de Hoog, R., Shadbolt, N., Van de Velde, W., and Wielinga, B., Knowledge Engineering and Management- The Common-KADS Methodology the MIT Press, 2000. Thomas W. M., Kevin C., and George A.H., Organizing Business Knowledge-The MIT Process Hand Book, MIT Press, 2003.